Stanford AI 课程地图 EN GitHub ↗
← 返回课程地图
Stanford · 2026—27 CS 227A 机器人、具身智能 ★★★★☆

核心课程 · 机器人、具身智能

机器人感知

Robot Perception (EE 227)

机器人感知:视觉/触觉/声学传感器 + 算法,在仿真里搭一套 vision-based 机器人系统。

3–4 学分 2026 秋 Letter or Credit/No Credit ★★★★☆
机器人感知视觉仿真

2026-27 秋季开设;官网暂未找到稳定链接。

课程定位

CS 227A(EE 227)由 Shuran Song 讲授,聚焦机器人如何从原始传感器数据中理解环境并做决策。课程覆盖传感器硬件、标定、视觉感知、状态估计和决策,作业会逐步搭建一个基于视觉的机器人系统。

主要内容

  1. 传感器:RGB-D、触觉、声学、IMU
  2. 标定:相机内参/外参、手眼标定
  3. 视觉感知:检测、分割、位姿估计
  4. 状态估计:滤波、融合
  5. 决策:从感知到动作
  6. 仿真实践
  7. 项目/报告(4 units 需额外报告)

先修

Python 编程;线性代数;数学基础会在课上复习,但编程要求高。

考核与工作量

一系列编程作业 + 可选 final report(4 units)。具体以当季官网为准。

适合谁

机器人感知、具身智能、3D 视觉方向。

学习建议

先修 CS 231N 或 CS 237A。作业编程量大,建议熟悉 PyTorch 和仿真环境。

相关课程

官方课程描述(英文原文)

Robot Perception is the cornerstone of modern robotics, enabling machines to interpret, understand, and respond to an array of sensory information they encounter. In the course, students will study the basic principles of typical sensor hardware on a robotics system (e.g., vision, tactile, and acoustic sensors), the algorithms that process the raw sensory data, and make actionable decisions from that information. Over the course of the semester, students will incrementally build their own vision-based robotics system in simulation via a series of homework coding assignments. Students enrolling 4 units will be required to submit an additional final written report. Prerequisites: This course requires programming experience in python as well as basic knowledge of linear algebra. Most of the required mathematical concepts will be reviewed, but it will be assumed that students have strong programming skills. All the homework requires extensive programming.